Geopolitical network

Introduction

One of the central problem of the worldregio project is to propose methods of regionalization based on the division of a network of states into sub-networks (components) according to a dual criteria :

  • geometric proximity : the components should fulfill a first condition of proximity in the graph
  • attribute proximity : the components should fulfill a condition of similarity, level of interaction, etc. which is supposed to be independent from their geometric proximity.

The definition of a geopolitical network at world level implies the joint definition of two lists of objects :

  • a finite list of political units called states
  • a finite list of relations between political units called borders

Notice that states and borders are used here in the abstract sense of edges and vertices which means that they do not necessary fit with the common sense idea.

A. GEOPOLITICAL VERTICES

Which list of units ?

The definition of a finite list of states will necessarily be complicated because many contested territories around the world are not recognized by all countries of the world and are therefore difficult to define in terms of geometry and are missing in the databases where the attributes used for regionalization are available.

Many list of countries are available in open source mapping packages but with different lists of countries and also different geometries !

The world of gadm

The map of the world provided by gadm is widely used as it provides data at different levels of administrative divisions and with different resolutions. We can therefore analyze what is proposed at the top level.

terra 1.7.78

Attaching package: 'terra'
The following object is masked from 'package:cartography':

    north
The following object is masked from 'package:knitr':

    spin

We find 231 territorial units that can be states member of United Nations but also non recognized territories .

  • Taiwan : the territory is relatively well defined but the country is not present in UN official databases because China consider it as a province of China. But in other databases (e.g. about trade), Taiwan exists.
  • Northern Cyprus : the territory is recognized as a state only by Turkey. The EU consider it as an occupied part of Cyprus.
  • Abkhazia, Southern Ossetia, Transnitria, … are de facto states recognized by Russia.
  • Kosovo : not recognized by all member states of EU

One of the most original unit propose by this database is Akrotiri and Dekalia which is a the territory of two military bases of UK located in the south of Cyprus.

Akrotiri and Dhekelia , officially the Sovereign Base Areas of Akrotiri and Dhekelia[2] (SBA),[a] is a British Overseas Territory on the island of Cyprus. The areas, which include British military bases and installations that were formerly part of the Crown colony of Cyprus, were retained by the British under the 1960 treaty of independence signed by the United Kingdom, Greece, Turkey and representatives from the Greek and Turkish Cypriot communities. The territory serves as a station for signals intelligence and is thereby part of the United Kingdom’s surveillance-gathering work in the Mediterranean and the Middle East.

Source : Akrotiri and Dhekelia, in English Wikipedia, 9 nov. 2024

The world of cshapes

A more precise and complete databases can be found in the pakage cshape which provide an historical database of geopolitical divisions between 1886 and 2019. This database has been established by researchers of the correlated of war project with the purpose of analyzig the origin of conflicts and trying to prevent them Schvitz et al. (2021).

We will firstly examine the situation of the world in 2019 according to this database.

This map is limited to 181 territorial units, from which 7 are considered as “colony” :

Territory declared as ‘colony’ by cshapes
country_name start end status owner capname
Puerto Rico 1898-12-10 2019-12-31 colony 2 San Juan
Guadeloupe 1886-01-01 2019-12-31 colony 220 Basse-Terre
Martinique 1886-01-01 2019-12-31 colony 220 Fort-de-France
French Guyana 1946-03-19 2019-12-31 colony 220 Cayenne
Reunion 1886-01-01 2019-12-31 colony 220 Saint-Denis
New Caledonia and Dependencies 1886-01-01 2019-12-31 colony 220 Nouméa
French Polynesia 1903-05-19 2019-12-31 colony 220 Pape’ete

We can notice that the number of territorial units is much smaller than in the case of gadm database. Many countries are missing as well as many overseas territories, in particular from UK. The ‘colony’ that are mentionned are in fact part of nation states (France, USA) that are allowed to participate to national elections and are in practice strongly linked to the nations they belong to.

The world of Natural Earth

As third and last example, we can also analyze the world presented in the famous package rnaturalearth and the complementary package rnaturaleathdata. The map provided by this package are produced by volunteers and coordinated by the North American Cartographic Information Society (NACIS).

Natural Earth is a public domain map dataset available at 1:10m, 1:50m, and 1:110 million scales. Featuring tightly integrated vector and raster data, with Natural Earth you can make a variety of visually pleasing, well-crafted maps with cartography or GIS software. Natural Earth was built through a collaboration of many volunteers and is supported by NACIS (North American Cartographic Information Society), and is free for use in any type of project (see our Terms of Use page for more information).

Source: Natural Earth website

This database is much more complex because they are a lot of option for the creation of the map, depending of the request on the type of unit and the scale of generalization. If we use the smallest level of resolution (10 meters) we can fo example obtain 298 territorial units.


Attaching package: 'rnaturalearthdata'
The following object is masked from 'package:rnaturalearth':

    countries110
name_long sovereignt type
Korean DMZ (north) North Korea Overlay
Brazilian Island Brazilian Island Indeterminate
Southern Patagonian Ice Field Southern Patagonian Ice Field Indeterminate
Bir Tawil Bir Tawil Indeterminate
Heard I. and McDonald Islands Australia Dependency
Bouvet Island Norway Dependency
Jarvis Island United States of America Geo unit
Baker Island United States of America Geo unit
Howland Island United States of America Geo unit
Midway Islands United States of America Geo unit

Which time period ?

Another problem is the changing list of states over the period of observation. We can use the another time the package cshapes to illustrate the changing divisions of the world in terms of geometry and political status.

Situation in 1948

Situation in 1968

Situation in 1988

Situation in 2018

Which geometry ?

Admitting that we have succeeded in the definition of a finite list of states, another problem will occur concerning the geometry of states which is generally not limited to a single polygon but can generally be defined as a multipolygon which can include very remote pieces of territory. According to the list of pieces of territory that we consider, the network of borders will be different and the results of the regionalization procedure can be heavily modified.

The case of Germany

Germany is an ideal situation where the state is made of a single polygon (except some nearby islands like Rügen) and where we can easily provide a list of 9 countries sharing a common terrestrial border (Denmark, The Nethetherlands, Belgium, Luxembourg, France, Switzerland, Austria, Czech republic, Poland).

The case of Denmark

The case of Denmark is more difficult because we have clearly a decision to take on the case of Greenland. Is it a part of Denmark or is it a separated country ?

Greenland is an autonomous country within the Kingdom of Denmark. Although Greenland is geographically a part of the North American continent, it has been politically and culturally associated with Europe for about a millennium. Since 1721, Denmark has held colonies in Greenland, but the country was made part of Denmark in 1953. In 1979 Denmark granted Home Rule to Greenland, and in 2009 expanded Self Rule was inaugurated, transferring yet more decision making power and more responsibilities to the Greenlandic government. Under the new structure, gradually Greenland can assume more and more responsibilities from Denmark, when it is ready for it.

Source : Visit Greenland )

In the case of aggregation of Denmark and greenland in a single spatial unit, it is interesting to observe that the definition of the centroid of the multipolygon would probably be located in the center of the inlandsis as the algorithm will generally choose the center of largest polygon

The case of France

France - as well as UK - are obviously very difficult cases because of the great number of so-called remote territories inherited from the colonial period. This remote territories are characterized by very different juridic status and it is impossible to adopt a common rule of decision for all of them.

Overseas France (French: France d’outre-mer, also France ultramarine)[note 3] consists of 13 French territories outside Europe, mostly the remnants of the French colonial empire that remained a part of the French state under various statuses after decolonisation. Most, but not all, are part of the European Union.

“Overseas France” is a collective name; while used in everyday life in France, it is not an administrative designation in its own right. Instead, the five overseas regions have exactly the same administrative status as the thirteen metropolitan regions; the five overseas collectivities are semi-autonomous; and New Caledonia is an autonomous territory. Overseas France includes island territories in the Atlantic, Pacific and Indian Oceans, French Guiana on the South American continent, and several peri-Antarctic islands as well as a claim in Antarctica. Excluding the district of Adélie Land, where French sovereignty is effective de jure by French law, but where the French exclusive claim on this part of Antarctica is frozen by the Antarctic Treaty (signed in 1959), overseas France covers a land area of 120,396 km2 (46,485 sq mi) and accounts for 18.0% of the French Republic’s land territory.Its exclusive economic zone (EEZ) of 9,825,538 km2 (3,793,661 sq mi) accounts for 96.7% of the EEZ of the French Republic.

Source : Overseas France in English Wikipedia, 9 nov. 2024

B. GEOPOLITICAL EDGES

The example of 2019

Map

We start our analysis by the choice of a reference map. We use here a map of 2019 that fit with the information of the trade database ‘gravity’ from the CEPII and which is made of 233 spatial units from which 194 are states and 39 are dependant territories. This map has been elaborated through a combination of cshapes and gadm sources.

Warning: st_centroid assumes attributes are constant over geometries

Contiguity I : Land borders

An apparently simple approach of the problem is to use land borders as criterium of definition of edges linking territories. The gravity database proposes a variable contig that seems apparently to fulfill perfectly our expectations.

Extract from the contiguity database of CEPII
iso3_o iso3_d contig
AFG CHN 1
AFG IRN 1
AFG PAK 1
AFG TJK 1
AFG TKM 1
AFG UZB 1
AGO COD 1
AGO COG 1
AGO NAM 1
AGO ZMB 1

We can produce a map of the number of border for each country of our database.

Warning: st_centroid assumes attributes are constant over geometries
Joining with `by = join_by(iso3)`
72 '0' values are not plotted on the map.

We observe that 66 countries or territory doesn’t have any land borders which creates isolated vertices in large parts of the world. The countries with the greatest number of land borders are Russia (16), China (14), Brazil (10), RD Congo and Germany (9). USA has only two borders, Canada only one and Australia zero.

Even if we eliminate the countries without any land borders, the resulting graph appears not connected and is divided in several components.

Contiguity II : Sea borders

Distance I : k nearest neighbours

An obvious solution to solve the problem of isolated units could be to use the nearest neighbor method i.e. to compute for each country or territory the distance with other country or territories and keep the k units that are at the lowest distance. Of course, this approach implies two choices that will have an influence on the results :

  • the choice of the measure of distance
  • the choice of the number of neighbors allocated to each country

It is also important to consider that the method is not symetrical because a country i can be the nearest neighbor of j but k is the nearest neighbor of i. So we have to introduce a third choice which is to transform the matrix in a symetric one or not.

As an example, we will extract from the CEPII database the distance between capital cities of each territory (distcap) and select the value k=5 which means that we will consider the five nearest neighbors.

Nearest neighbors based on CEPII distance between capitals
iso3_o iso3_d distcap rnk
ABW CUW 128 1
ABW BES 196 2
ABW VEN 411 3
ABW DOM 660 4
ABW HTI 710 5
AFG PAK 370 1
AFG TJK 446 2
AFG UZB 748 3
AFG IND 1002 4
AFG KGZ 1038 5

We proceed now to the symetrisation of the matrix which will increase the number of links of some states to a value greater than 5.

We can produce a map of the number of links for each country of our database.

Warning: st_centroid assumes attributes are constant over geometries
Joining with `by = join_by(iso3)`

The distribution of neighbors is more regular with a minimum of 5 (our initial choice) and a maximum of 10 (due to the symetrization of the matrix).

The resulting network is made of a single component which was not necessarily an obvious result. With a lower value of K, we would have probably obtained a more fragmented network with isolated components in Pacific or separation between the old and the new continents (that are connected via Greenland).

From a geopolitical point of view, it is interesting to observe that the network is organized in strongly connected components linked by bridges that are country with high level of betweeness like Turkey, Mali, Greenland … It should be possible to demonstrate that those countries with high level of betweeness are generally strategic places where crises and conflict are likely to take place. But of course this result has to be balanced by the fact that the different nodes does not have the same weight in international affairs and we can propose alternative methods taking into account the power of countries and the intensity of flows that link them.

Delaunay-Voronoï tesselation

An alternative family of solution is the creation of a Delaunay-Voronoï tesselation based on the edges corresponding to the centers of states.

If we consider that the earth is “flat”, the problemis very simple because we just need to put the \((x,y)\) coordinates of the center of states and apply some functions of the spdep package. We obtain in a few seconds the list of states linked to each edge :

Warning: st_centroid assumes attributes are constant over geometries
Loading required package: spData
Joining with `by = join_by(i)`
Joining with `by = join_by(j)`
Liste des liens (non pondérés)
i j iso3i iso3j
1 3 USA CAN
1 4 USA BHS
1 5 USA CUB
1 13 USA MEX
1 198 USA KIR
1 220 USA BMU
2 7 PRI DOM
2 185 PRI BES
2 189 PRI DMA
2 199 PRI KNA

We can then visualize the graph on a world map of states :

The result is interesting because they are no more isolated states and some links are crossing the seas and ocean in order to cover all the world. But some issues can be pointed.

  1. Very little territories can obtain a dramatic importance when they are isolated island in the middle of oceanic area. For example the small islands of Saint Helena (SHN) , Cabo Verde (CPV) and Saint-Pierre et Miquelon (SPM) appears as gatekeepers between Americas and Europa or Africa.

  2. The tesselation has been made on a planar surface which exclude links across the Pacific Ocean. If we had chosen the same projection but with another longitude of reference, we would have observed links across the Pacific, linking America and Asia.

Warning: st_centroid assumes attributes are constant over geometries
Joining with `by = join_by(i)`
Joining with `by = join_by(j)`

  1. The result of this method is strongly from the location of the center of the states chosen (as it was the case for the k-nearest neighbour method). We have chosen here the center of the largest polygon but we would have obtained different results if we had chosen the capital city (Cf. the case of Russia).

As a whole, the method is interesting but should be improved in different ways :

  1. Computing Delaunay-Voronoï directly on a spherical object. Our research on internet reveals that many authors has proposed solution to this problem with applications in the languages Python or Observable. You can for example have a look at d3-geo-voronoi application.

  2. Better results could probably be obtained by the use of multiple centers inside each state. As an example you can look at a map of triangulation of world airports which could be eventually aggregated at national level in order to obtain an international network.

Spherical Delaunay-Voronoï triangulation applied to the distribution of world airports

References

Conte, Maddalena, Pierre Cotterlaz, and Thierry Mayer. 2022. “The CEPII Gravity Database.” https://www.cepii.fr/PDF_PUB/wp/2022/wp2022-05.pdf.
Schvitz, Guy, Luc Girardin, Seraina Rüegger, Nils B. Weidmann, Lars-Erik Cederman, and Kristian Skrede Gleditsch. 2021. “Mapping the International System, 1886-2019: The CShapes 2.0 Dataset.” Journal of Conflict Resolution 66 (1): 144–61. https://doi.org/10.1177/00220027211013563.